Accession Number : AD0672089

Title :   LEARNING CONTROL SYSTEMS AND PATTERN RECOGNITION.

Descriptive Note : Semi-annual progress rept. no. 1 on project THEMIS, 15 Sep 67-15 Mar 68,

Corporate Author : VIRGINIA UNIV CHARLOTTESVILLE RESEARCH LABS FOR THE ENGINEERING SCIENCES

Personal Author(s) : Moore,James W. ; McVey,E. S. ; Ramey,R. L. ; White,E. J. ; Bowman,R. M.

Report Date : 15 APR 1968

Pagination or Media Count : 125

Abstract : The report covers progress in the first six months on a Project THEMIS Contract for research in Learning Control Systems and Pattern Recognition. Topics covered include the general approach to learning control systems, finite state sequential machines, thin film deposition research, adjustment of light gain of phototransistors, and some considerations on the optimal design for a single font character recognition machine. This latter is in the form of a final report and is described below in more detail. A means of constructing an inexpensive single font alphabetic character recognition machine is presented. Major parameters are considered and constraints chosen to minimize costs. A method for choosing a minimum set of invariant characteristics using probability theory and a priori noise immunity is presented. Results include a relation between noise immunity and the array size into which characters are quantized, indicating a maximum noise immunity of 30.9%. A hypothesis relating character set size and character similarities is stated. (Author)

Descriptors :   (*READING MACHINES, *PATTERN RECOGNITION), CHARACTER RECOGNITION, ADAPTIVE SYSTEMS, COMPUTER LOGIC, PHOTOSENSITIVITY, INTEGRATED CIRCUITS, VAPOR PLATING, PHOTOELECTRIC CELLS(SEMICONDUCTOR), VACUUM APPARATUS, LEARNING MACHINES, COMPUTER PROGRAMS, FEASIBILITY STUDIES

Subject Categories : Computer Programming and Software
      Computer Hardware

Distribution Statement : APPROVED FOR PUBLIC RELEASE